A festive game of Scrabble is a time-tested method of surviving the extended company of obnoxious family members. But losing to a crabby relative can make their company even worse.
But this year, uncle Nigel (name changed to protect identity) will face a different challenge. Thanks to a new AI version of the classic board game, his distressing knowledge of the dictionary will be of no use at all — because real words no longer count.
The BLABRECS system is the brainchild of Max Kreminski, an AI researcher and game designer. He describes his creation as “like Scrabble but worse.”
The browser-based tool is designed to run alongside a regular Scrabble game. But under the new rules, you can only play words that the AI says don’t exist — but sound as if they could.
After arranging your tiles into an order that looks like gibberish, enter the word into the BLABRECS text box to check whether the AI deems it appropriately nonsensical. If it’s accepted, hit the “Play It” button to add the word — and your definition for it — to the lexicon.
The system checks whether a word is meaningless by running it through a Markov model trained on the ENABLE list of more than 173,000 words, which is used as a reference dictionary for numerous word games.
“It looks at the statistical patterns of letter sequences in English words and uses this information to determine how likely a sequence of letters is to be a real English word,” Kreminski explains on the BLABRECS website. “Then it rejects both real dictionary words and fake words that it deems insufficiently plausible.”
— Brian Mastenbrook (@bmastenbrook) December 14, 2020
Kreminski admits that the AI sometimes accepts real words as gobbledeegook, particularly inflected forms of base words and proper nouns that Scrabble generally prohibits.
When this happens, he suggests playing by the spirit of BLABRECS — whatever that means to you:
Is the game primarily about exploring the vast ‘shadow English’ implied by the statistical distribution of letter sequences, or is it about the inherent absurdity of an external authority presuming to dictate your language to you? Either interpretation seems valid to me.
Kreminiski is also considering adding AI opponents to future versions of the system, so you wouldn’t have to interact with toxic relatives at all. Roll on Christmas 2021.
Get the Neural newsletter
Greetings Humanoids! Did you know we have a newsletter all about AI? You can subscribe to it right here.Follow @neural